Getting started with AI, one step at a time

How to use this path

This path is for you if you’ve tried ChatGPT or another AI assistant without really knowing how to get the most out of it, or if you’ve never dared to try at all. The steps build on each other, so it’s best to follow them in order. Go at your own pace, one step per session if you like, and come back whenever something feels unclear. There’s nothing to sign up for, and the tools we suggest run entirely in your browser: nothing you type is sent anywhere.

The 12 steps

Step 1 — Understand what artificial intelligence is

Artificial intelligence refers to software that can handle tasks we used to associate with human intelligence: understanding a sentence, recognizing an image, summarizing a text. The idea isn’t new: Alan Turing proposed his famous test in 1950, and the term itself was coined at the Dartmouth conference in 1956. What changed recently is that conversational tools became available to everyone, a shift made popular by the launch of ChatGPT in November 2022.

→ To do: browse the AI timeline to see the major milestones, from the first programs to today’s assistants.

Step 2 — How an assistant like ChatGPT actually works

An assistant like ChatGPT runs on a language model: a program trained on huge amounts of text to predict, piece by piece, the most likely continuation of a sentence. It doesn’t “know” things the way you do; it produces plausible text based on patterns it has learned. That’s why it sounds so fluent, and also why it sometimes gets things wrong. Several companies build these models, each with its own technical and business choices.

→ To do: find out who makes these models on the AI labs page.

Step 3 — The words worth knowing

A handful of terms come up everywhere. The model is the software engine behind the assistant. A prompt is your request. A token is a small chunk of text the model processes, often part of a word. The context is everything the assistant keeps in mind during the conversation. An AI agent is an AI that takes actions, not just one that answers. Knowing these words keeps you from feeling lost in settings and articles.

→ To do: keep the AI glossary handy and come back to it whenever a word trips you up.

Step 4 — Pick your assistant

ChatGPT isn’t the only option: other assistants exist, each with different strengths (writing, web search, integration with your phone or email). Most of them offer a free plan, which is plenty for beginners. The best AI for you depends on what you want to do with it every day, not on which brand is talked about the most.

→ To do: take the “Which AI is right for you?” quiz, then compare plans in the AI assistants comparison.

Step 5 — Write a clear request

The quality of the answer depends heavily on the quality of the question. A good prompt states the role you expect (“act as a travel planner”), the background, the result you want and the format (bullet list, table, short email). Feel free to steer the conversation as you go: “shorter,” “simpler,” “with an example.” An assistant never gets offended when you ask it to try again.

→ To do: build your first structured request with the prompt generator.

Step 6 — Why AI sometimes makes things up, and how to check

Sometimes an assistant confidently states something false. This is called a “hallucination.” It isn’t a deliberate lie; it’s a side effect of how the model generates plausible-sounding text. For anything that matters (health, money, official paperwork, dates, quotes), check with a reliable source. You can also ask the assistant where the information comes from, then open those sources yourself.

→ To do: next time an answer really matters, verify at least one key fact on an official website before relying on it.

Step 7 — Privacy: what not to share

What you type into an assistant may be stored by the company behind it and, depending on your settings, used to improve its models. So avoid pasting passwords, credit card numbers, ID documents, medical records or your employer’s confidential information. When you ask for help with a personal situation, swap real names for made-up ones: the answer will be just as useful.

→ To do: open your assistant’s settings and find the option that controls whether your conversations are used for training.

Step 8 — Everyday uses

This is where AI really pays off: drafting a tricky email, writing a letter to cancel a subscription, summarizing a long document, planning a trip, coming up with dinner ideas from what’s left in the fridge. The habit to build: give concrete details (dates, tone, recipient, constraints) and reread before sending. The assistant writes the first draft; you make the final call.

→ To do: pick a real task from your week and prepare the request with the prompt generator.

Step 9 — Talk to AI with voice mode

Several assistants let you have a spoken conversation from their mobile app, a bit like a phone call. It’s handy while cooking, walking, or practicing a foreign language. The same rules apply as in writing: be specific, don’t say anything sensitive out loud, and be wary of answers that sound a little too sure of themselves.

→ To do: if your assistant offers voice mode, try asking for a recipe or rehearsing a conversation in another language; if not, compare features in the AI assistants comparison.

Step 10 — Spot AI-powered scams

Scammers use AI too: fake apps posing as well-known assistants, a cloned voice of a loved one urgently asking for money, doctored videos (known as “deepfakes”) of celebrities promoting a miracle investment. A few habits go a long way: download apps only from official app stores, call people back on their usual number, and be suspicious of any urgency or promise of easy money.

→ To do: agree on a family code word with your loved ones to ask for if you ever get a suspicious call.

Step 11 — What is an AI agent, and what can it do for you?

An AI agent goes further than an assistant that answers questions: it can act on your behalf across several steps, such as searching for tickets, filling out a form, sorting your inbox or preparing an order. That’s convenient, but an agent’s mistakes have real consequences: an unwanted purchase, a message sent to the wrong person. It’s wise to start with low-stakes, reversible tasks and keep the final approval for yourself.

→ To do: describe a task you’d like to hand off in the “Can I trust an agent with this?” tool before you try it.

Step 12 — Take stock, then move on

You now have the basics: understand, choose, ask, check, protect yourself. It’s time to see where you stand, with no pressure: this isn’t an exam, just a benchmark for yourself. Depending on your result, you can review a step or move on to the next level, which covers advanced uses and consumer AI agents.

→ To do: take the AI level test, then continue with the Go further path.

After the path

Being independent doesn’t mean knowing everything. It means knowing when AI saves you time, when it’s likely to be wrong and when you’re better off without it. Models and plans change fast, so the site refreshes its data every week, with sources and dates, to help you keep up without effort. To see which models stand out right now, the weekly AI ranking is a great place to start.